Evaluation of BAYESIL for automated annotation of 1H NMR data using limited sample volumes: application to African elephant serum

Metabolomics (2023) 19:31 https://doi.org/10.1007/s11306-023-02001-1 ORIGINAL ARTICLE Evaluation of BAYESIL for automated annotation of 1H NMR data using limited sample volumes: application to African elephant serum Christiaan De Wet van Zyl1 · Mari van Reenen1 · Gernot Osthoff2 · Ilse du Preez1 Received: 14 October 2022 / Accepted: 22 March 2023 © The Author(s) 2023 Abstract Introduction Technological advancements enabled the analyses of limited sample volumes on 1H NMR. Manual spectral profiling of the data is, however, complex, and timely. Objective To evaluate the performance of BAYESIL for automated identification and quantification of 1H NMR spectra of limited volume samples. Method Aliquots of a pooled African elephant serum sample were analyzed using standard and reduced volumes. Performance was evaluated on confidence scores, non-detects and laboratory CV. Results Of the 47 compounds detected, 28 had favorable performances. The approach could differentiate samples based on biological variation. Conclusions BAYESIL is valuable for limited sample 1H NMR data analyses. Keywords BAYESIL · 1H NMR · Limited sample volume · Serum · Elephant 1 Introduction Metabolomics is defined as the measurement of the changes in the low molecular weight compounds (metabolites) in a specific biological specimen because of internal fluctuations and/or external disturbances (Lenz & Wilson, 2007). Metabolomics is well established in animal studies including species such as fish (Tavares et al., 2022), mussels (Boaz et al., 2012; Wu & Wang, 2010), mice (Mason et al., 2018), rats (Maulidiani et al., 2017) seals (Boaz et al., 2012) and more. For this reason, researchers are striving Ilse du Preez Christiaan De Wet van Zyl Mari van Reenen Gernot Osthoff 1 Centre for Human Metabolomics, North-West University, Potchefstroom, South Africa 2 Department of Microbiology and Biochemistry, University of the Free State, Bloemfontein, South Africa to optimize and standardize protocols for the analyses of biological specimens collected from specific animal species, such as was recently done for the proton nuclear magnetic resonance (1H NMR) analyses of whole blood (Lenz & Wilson, 2007; Wood et al., 2022) and milk (Osthoff et al., 2023) collected from African elephants. The use of 1H NMR as an analytical tool for metabolomics analysis has many advantages, including the fact that it requires minimal, low cost sample preparation, and is robust, rapid, unbiased, reproducible, non-selective, non-destructive and quantitative (Lenz & Wilson, 2007). This technique, nevertheless, has a low sensitivity when compared to other methods, such as mass spectroscopy, and therefore typically requires larger sample volumes. Advancements in the hardware technology, and the development of systems such as the MATCH adapter, have, however, facilitated the optimization of limited sample volume analyses (Mason et al., 2018). A drawback of 1H NMR in untargeted metabolomics studies, is the high complexity of spectral profiling, specifically the annotation and quantification of metabolites from spectra. The processing is time consuming, labor intensive, requires expert skills and is subjective and prone to human error (Tredwell et al., 2011). BAYESIL was developed as an automated alternative to manual 1H NMR spectral profiling (Ravanbakhsh et al., 2015). The freeware automatically 13 31 Page 2 of 7 C. D. W. van Zyl et al. quantifies metabolites, which are identified with specified confidence based on a match to a built-in library, in less than 5 min per spectrum (Ravanbakhsh et al., 2015). This open source software has been evaluated and deemed useful for metabolomics applications by various groups. Such assessments typically entail comparisons between BAYESIL and manual spectral profiling (Tavares et al., 2022), or between BAYESIL and other automated software tools (Lipfert et al., 2019; Maulidiani et al., 2017). In these studies, the prescribed sample volumes and preparation methods are used. For serum this includes filtering, buffering and the addition of reference standards. It is suggested that a total serum volume, after filtration, of 540 µL is used per sample. The aim of our study was to evaluate the performance of BAYESIL when using a reduced sample volume of 54 µL, after filtration (10% of that stipulated in the protocol). The ability of the miniaturized approach to differentiate between sample groups was subsequently explored by graphically presenting metabolite differences between serum collected from female (lactating and using contraceptives) and male African elephants and by evaluating their biological relevance. Outputs from the study would give an indication of the sensitivity that can be expected when spectra are processed by a non-expert, and would stipulate the feasibility of automated, high-throughput, limited sample volume, 1H NMR analyses. To the best of our knowledge, this is the first evaluation of the freeware using very low sample volumes. In addition, although BAYESIL has been used as a tool in various studies using both human (Garg et al., 2018; Grimaldi et al., 2018; Maulidiani et al., 2017) and animal samples (Tavares et al., 2022), this study is the first to assess its value in African elephant research. into an 8.5ml yellow top vacuum tube. The tubes were kept on ice for 30 min during transportation to the laboratory, and the serum was frozen at -80 ℃ until further analysis. 2 Materials and methods 2.4 Standard sample preparation 2.1 Animals and sample collection For the standard sample preparation, the method suggested by BAYESIL was applied. Instead of the 3 500 molecular weight cutoff MWCO centrifugal units recommended by BAYESIL, 10 000 molecular weight cutoff (MWCO) centrifugal units, which are readily available in our laboratory, were used. Of each QC aliquot, 800 µL was added to an Amicon Ultra-2mL 10 000 MWCO centrifugal unit (5 x pre-rinsed with water) and centrifuged at 2860 x g for 40 min. From the filtrate, a volume of 540 µL was added to a microcentrifuge tube that contained 60 µL of the 1H NMR buffer solution (9:1 ratio). After a brief vortex step, 600 µL of the buffered filtrate was transferred to a 5 mm 1H NMR tube and sealed with a cap. Blood samples were obtained from two female lactating African elephants (Loxodonta africana) (1) Mussina at 15, 16, 16.3, 16.5, 17.3 and 17.7, and (2) Shan at 27.5, 27.9, 28.1, 29.1 and 29.4 months of lactation. In addition, blood was also collected at six time points, roughly 2 weeks apart, from the same two elephants (after lactation), from a third female, Naledi, and two males, Chova and Chishuru. The three non-lactating female animals were on hormonal contraceptives at the time of sample collection. The elephants roamed free in the Adventures with Elephants Reserve (Bela Bela, Limpopo province, South Africa). The nutrition, care and well-being of the animals were described in a previous paper (Kobeni et al., 202 (...truncated)